Inferring the Invisible: Recurrent Neuro-Symbolic Forward Chaining Network
Keywords: Neuro Symbolic Reasoning
Abstract: A key challenge in artificial intelligence is inferring underlying factors that are not directly observable but are crucial for understanding and predicting complex behaviors. In this paper, we introduce a novel neural-symbolic framework that advances beyond traditional rule induction by integrating latent predicate discovery with rule learning. Our approach utilizes a recurrent unit to iteratively refine and learn rules from observed data, employing dynamic programming techniques to identify fixed points and solve complex problems. This framework enables the discovery of hidden predicates—such as user engagement or underlying motivations—that influence observable outcomes but are not directly grounded in the data.By encoding both explicit and latent predicates into a unified rule embedding, our method facilitates a deeper understanding of complex phenomena and enhances predictive accuracy. This joint learning process captures explicit relationships and invents new predicates essential for comprehensive inference. We validate our method across various tasks, demonstrating its capability to reveal hidden structures and enhance symbolic reasoning with deeper, more accurate insights.
Supplementary Material: zip
Primary Area: neurosymbolic & hybrid AI systems (physics-informed, logic & formal reasoning, etc.)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 2758
Loading